The pollen imprint of the Juniperus, Picea and

Juglans forest belts of Kyrgyzstan, Central Asia

Ruth Beer1, Willy Tinner1, Gabriele Carraro 2, Ennio Grisa 3

1Section of Palaeoecology, Istitute of Plant Sciences, University of Bern, Altenbergrain 21,

CH 3013 Bern, Switzerland

2 Dionea SA, Lungolago Motta 8, CH 6600 Locarno

3 Intercooperation KIRFOR (Kyrgyz-Swiss Forestry Support Programme), Kyrgyzstan,

Karagashova Rosha 15 720000 Bishkek, Kyrgyz Republic

Corresponding author: [email protected] Abstract Surface pollen deposition at three different sites and forest belts (Juniperus, Picea, and Juglans forests) in Kyrgyzstan have been investigated for the first time in order to assess the relationship between the modern vegetation and the pollen composition with emphasis on tree-crown density. Our results show a close relationship between vegetation and pollen for the Juniperus (e.g. r2 = 0.76 between crown density and arboreal pollen) and Picea (r2 = 0.85) sites, whereas the linkage is weaker at the Juglans sites (r2 = 0.35) and in mixed forests (r2 = 0.32). Our investigation may help to better reconstruct the Quaternary vegetation and climate history of these Central Asian vegetational belts.

Key Words: surface samples, tree crown cover, palynology, Kyrgyzstan

Introduction Much of the uncertainty in the interpretation of Late Quaternary pollen assemblages can be removed by the judicious use of pollen surface samples from a variety of vegetational formations. The technique is especially useful in new regions of investigation, where not much is known of the vegetational imprint on the pollen sequence (Wright, 1967). A special focus is the separation of pollen sources from local to regional scales (Wright, 1967). In Europe, such efforts were undertaken already more than fifty years ago to better understand the linkages between vegetation and pollen (e.g. in treeline situations Firbas, 1934; Welten, 1950) More recently, several studies on pollen dispersal, pollen source area and pollen productivity estimates have been carried out in Northern America (Janssen, 1967; Calcote, 1995; Jackson and Kearsley, 1998) and in Scandinavia where modelling studies are used to calibrate pollen and vegetation cover (Sugita, 1994; Sugita et al., 1999; Bunting et al., 2004; Broström et al., 2004, 2005.) Near our study region, Wright et al. (1967) studied the relationship of modern pollen rain and plant cover in western Iran. Further El-Moslimany (1990) investigated sites in the Middle East. Kvavadze and Stuchlik (1990) studied the sub- recent spore and pollen spectra and their relation to recent vegetation belts in north-western Georgia. Liu et al. (1999) made a study of the woodland ecotone in relationship to surface pollen in the Inner Mongolian Plateau, China. Recently a study was made on the quatitative relationship between modern pollen rain and climate in Tibet (Shen et al., 2006). However, so far no investigation on modern pollen rain in relation to vegetation has been made in Kyrgyzstan, Central Asia. The present work is useful to better interpret the Quaternary vegetation and climate history of this region.

Material and Methods A total of 43 surface samples have been collected at four different sites in Kyrgyzstan in three different vegetation types during two coring expeditions in 2003 and 2005 respectively. The samples of moss polster or loose surface soil were collected along transects, the dominant tree species and an estimation of the crown densities were noted. Vegetation releves were made on plots of 10 x 10 m around the surface soil sample and the species were listed in the order of their relevance. Kichiköl (N 39°59’16.4“ and E 073°33’04.0“) is a lake located at 2554 m a.s.l. on the northern slope of the Alay range (Fig 1). The modern forests consist of Juniperus turkestanica Kom. and J. semiglobosa Rgl. They are restricted to the north facing slopes and are rather open. Pastures and meadows occur in open patches of the forest and on south facing slopes.

2 Karaköl (N 42°50’20.2” and E 077°23’33.8”) is a subalpine lake located at 2353 m a.s.l. on the south-facing slope of the Kungey Alatau. Dense stands of Picea schrenkiana F. et M. form the forests in this area. Pastures and meadows are interspersed in the forest patches. The Bakaly lake (N 41°86’71.5’’ and E 071°96’01.1’’) lies south of the Chatkal range at 1880 m a.s.l. in the Sary Chelek reserve at the upper dispersal limit of Juglans regia L. The forests consist of a mosaic of Juglans regia L., Juniperus turkestanica Kom., J. semiglobosa Rgl., Picea schrenkiana F. et M., Abies semenovii Hill., Acer turkestanica Pax., Cerasus mahaleb (L.) Mill., Malus kirghisorum Al. et An. Theod., Prunus ssp., and Crataegus ssp. and are usually rather open Nischnie and Vierschnie Ozira are two lakes located in the Arslan-Bob region on the south- facing slope of the Fergana range (N 41°18’32.8 and E 072°57’52.6 and N 41°18’31.5” and E 072°57’30.3” respectively). The Nischnie lake lies at 1371 m a.s.l. and Vierschnie lake at 1440 m a.s.l. Both lakes lie in the so called walnut-fruit forest region. The local forests around the lakes are rather dense. Forests with Juglans regia L. together with Crataegus turkestanica A. Pojark., Malus kirghisorum Al. et An. Theod., Prunus mahaleb, Prunus sogdiana Vass., Cerasus mahaleb (L.) Mill. are predominant on north-facing slopes. Stands of Acer turkestanica Pax. together with the above mentioned tree species out of the Rosaceae are found on south-facing slopes while Pistacia vera forms thickets on the driest patches. For a thorough description of the walnut-fruit forests see Blaser et al. (1998) and Epple (2001). Some of these forests are under heavy grazing pressure as the cattle are rich in number and have free entry into the forests. Other human impact involves hay making, over-harvesting of nuts and uncontrolled timber exploitation (Blaser et al., 1998; Kolov, 1998; Venglovsky, 1998). At present, efforts are made to protect the forests of this region (Blaser et al., 1998).

Surface soil samples (moss polsters, top 2 cm of Ah-horizons) were treated with KOH and sieved with a mesh of 1mm then treated with HCL, HF and acetolysis (Moore et al., 1991). The pollen samples were suspended in glycerine. In each sample at least 600 pollen grains were counted and identified using pollen keys (Moore et al., 1991; Beug, 2004) pollen atlases (Reille, 1992; 1998) and the reference collection of the Institute of Plant Science, University of Bern. The pollen of Cyperaceae and spores were excluded from the pollen sum. On the basis of this sum, percentages of pollen types have been calculated. Sketches of the site were drawn according to the vegetational surveys along the axis of the pollen diagrams to visualize the landscape and the vegetation. Altitudinal variation was integrated but the distance to the lake is not drawn in linear measure.

Results Kichiköl: Juniperus forests a) crown densities (KKO1) Five samples were taken and analysed spanning from single Juniperus trees (sample 1, crown density 10 %) to a closed Juniperus forest (sample 5, crown density 100 %). Our results (Table 1, Fig. 2; 3) show that vegetation and pollen are in very good agreement, e.g., with increasing arboreal pollen (AP) values corresponding to increasing tree cover. Sample 1 and 2 represent Juniperus tree crown densities of 10 and 20 % respectively, both show Juniperus pollen values around 25 % in association with low stomata finds. Sample 3 represents a crown density of 45 % and yields a percentage of Juniperus pollen of 35. Juniperus stomata are found but with few counts only. In sample 4 a crown density of 80 % is reflected by 80 % Juniperus pollen and high stomata finds. In the vegetational plot 5 dense stands of Juniperus turkestanica and J. semiglobosa form a closed canopy. They are represented by 65 % Juniperus pollen in association with numerous stomata finds in sample 5. Single pollen grains of Prunus and Sorbus type pollen are found, but can not be related to trees growing in the

3 vegetational plots. Artemisia shows increasing values up to 45 % with diminishing crown densities of the Juniperus forest although not present locally in the releves. Chenopodiaceae pollen percentages reach a 10 % value only in plot 1. The Poaceae pollen reaches 15 % in sample 2 and 10 % in sample 1 and 3. High percentages of Cyperaceae pollen are recorded in samples 1 to 3, in part this may be related to the vicinity of riparian vegetation around the lake. b) transect Juniperus forest to meadows (KKO2) Five samples were taken along a transect from dense Juniperus stands (samples 1 and 2) on the north facing slope of the lake to the meadows on the south facing slopes (samples 3, 4 and 5). Fig. 4 shows the decreasing Juniperus pollen percentages with increasing distance to the forest. In the vegetational plot 1 Juniperus turkestanica and J. semiglobosa form dense stands with a crown density of 75 % on the steep slope above Kichiköl. Juniperus pollen reaches 35 % and numerous stomata are found. In plot 2 Juniperus trees form open stands with 30 % crown cover. Here the values of Juniperus pollen drop to 20 %, without any stomata finds. In the samples 3 to 5, Juniperus pollen reaches 10 %. Artemisia pollen is present throughout the transect with 45 to 55 %. The Artemisia pollen may originate from the local plants in the meadows. The pollen of Chenopodiaceae is represented by low but constant values around 5 to 10 %. Except for sample 1 the pollen of Poaceae show constant values around 10 %. Cichorioideae pollen type shows a peak of 10 % in sample 3, whereas the Mentha pollen type is found in high amounts in sample 4. Both types may mirror the meadows and pastures around the lake on the south facing slopes. Interestingly, the pollen of Cyperaceae show not only a peak in sample 3 near the lake but also in sample 5. This is easily explainable by the fact that Cyperaceae such as Carex are growing in both, upland and wetland communities.

Karaköl: Picea forest (KAO) Six samples were taken along a N-S transect at Karaköl starting in closed Picea forest stands on the north facing slope to the meadows around the lake and into the forests of the south facing slope (Fig 5). Fig 6 shows the very good agreement of the pollen percentages with tree crown cover. Closed forest stands of Picea schrenkiana with a crown density of 100 % are recorded in the vegetational plots 1 and 6. A crown density of 75 % is found in plot 5 whereas plots 4 and 2 represent semi-open stands. The pollen percentages in association with high finds of Picea stomata reflect the closeness of the stands with 80, 70 and around 50 % of tree cover, respectively. Sample 3 was taken in a meadow without Picea schrenkiana trees. It still shows a value of 30 % Picea pollen. Juniperus pollen is recorded throughout the transect, probably deriving from the Juniperus stands on the south-facing slopes in the area. A slightly higher percentage value is found in sample 4 where a sigle Juniperus tree is recorded locally. Steppe vegetation represented by Artemisia and Chenopodiaceae pollen is not present in the local vegetation but shows constant percentage values of 10 – 20 and < 10 % respectively, probably again originating from the south-facing, drier slopes. In sample 3 high percentage values of Asteroideae, Cichorioideae, Ranunculus acris type and a 10 % value of Poaceae pollen reflect local vegetation of meadows and pasturelands. The pollen of Cyperaceae type may mirror both the meadows and the riparian vegetation.

Bakaly: Mixed forests a) N-S transect (BAO2) Six samples were taken and analysed from a N-S transect in the mixed forests around Bakaly (Fig. 7). Generally, more tree species form more complex woodland communities than in the almost mono-specific southern Juniperus and northern Picea forests. Juniperus turkestanica and J. semiglobosa are present in plot 1 together with Crataegus turkestanica, Malus kirghisorum and Cerasus mahaleb. They build communities with a crown density of 70 % but

4 AP only reaches percentages of 25 % and is dominated by Juniperus. Though no trees are recorded in the vegetational plots 2 and 3 AP still accounts for 25 % in sample 2, but then drops to 10 % in sample 3. In plot 6 where Juniperus turkestanica grows together with Picea schrenkiana the sum of tree pollen reaches 30 % and is dominated by Juniperus and Picea. Picea schrenkiana grows in plot 5 and 6, but high pollen values which are associated with finds of Picea stomata are only reached in sample 6, where the crown density attains 80 %. Juglans pollen is present with low values in all the samples, although not growing in the vegetational plots. This pollen most likely originates from extra-local trees in the lake catchment. (see W-E transect BAO, Fig 8). Cerasus mahaleb, Malus kirghisorum and Crataegus sp. grow in plot 1. Malus kirghisorum and Crataegus turkestanica are found in plot 4 as well as Cerasus mahaleb in plot 6. The pollen of Sorbus type may partly represent these Rosaceae trees with single grains. Acer turkestanica grows in plot 6 but no pollen is found, however, single finds of Acer pollen occur in the samples 1 and 4. Pollen of taxa growing in the meadows and steppes such as Artemisia and Chenopodiaceae is present throughout the transect, with values around 15 and 10 %, respectively, although it is locally missing in the vegetational plots. The Poaceae pollen shows its lowest percentage value (10 %) in sample 6 where the crown density of the forest stand with Picea schrenkiana, Acer turkestanica, Cerasus mahaleb and Juniperus turkestanica reaches maximum values of 80 %. In plot 3 with open vegetation they reach 30 % along with high values of Cyperaceae. Plantago sp. though present in this sample with high pollen values is not found in the local vegetational survey. b) W-E transect (BAO) Ten samples were taken along a W-E transect at Bakaly integrating the different types of mixed forests with different crown covers (Fig. 8). Juniperus turkestanica and J. semiglobosa are present in the vegetational plots 1, 3, 6, 8, 9 and 10 where they form semi-open stands together with other tree species. Except for sample 5, the Juniperus trees are represented by percentage values from 10 to 40 percent. Although not present in the local vegetation releves the highest percentage value of Juniperus pollen is found in sample 7 (40%). In sample 6 they reach only 15 % in association with high finds of stomata. Juniperus turkestanica and J. semiglobosa form an open stand together with Malus kirghisorum in this vegetational plot. Betula pollen reaches 40 % in sample 4 and another 15 % in sample 5 in association with high finds of Cyperaceae pollen. It mirrors a stand of Betula sp. growing near the shore on upland soils at sampling site 4. Juglans regia is found in the plots 1, 8 and 9 but only in sample 1 it reaches higher values (20 %). As the forest stands are open (crown density 30 to 40 %), with only single Juglans regia trees, a higher relevance of Juglans pollen in sample 8 and 9 might be masked by the conspicuous presence of Poaceae pollen (55 and 40 % respectively). Picea pollen is constantly found throughout the transect with rather low values (< 10 %), though not present in the vegetational releves. The Sorbus type pollen is found in all the samples, except for 8 and 10. Rosaceae shrub and tree species are found throughout the transect, except for those samples near the lake shore (samples 4 and 5). Acer turcestanica is found in the vegetational plots 1, 3 and 8 but doesn’t show up in the pollen samples. Artemisia and Chenopodiaceae pollen are constantly represented in the pollen with 10 to 20 % and 3 to 10 %, respectively, although locally not recorded in the vegetation plots.

Nischnie and Vierschnie Ozira: Juglans forests (NOVO) Eleven samples were taken around the two lakes with the aim to cover the different local vegetation types. Juglans regia predominantly forms the forests in this region but thickets of several species of Rosaceae and stands of Acer are interspersed as well (Table 1). Fig. 9 shows that Juglans regia dominates the AP throughout the transect. The highest percentage value (85 %) of the species is found in sample 3, where it constitutes a mono-specific stand in the vegetational plot. Sample 3, 6, 9 and 11 show Juglans pollen values around 70 %, where

5 Juglans regia forms mixed stands. In samples 2, 4, 5 and 7 the pollen percentages attend 35 – 40 %, although Juglans being locally present only in the plots 5 and 6. Sample 10 shows a value of 18 % and sample 8 only 10 % with Juglans regia being absent in both plots. Juniperus pollen is present throughout the pollen spectrum with percentage values from 5 to 12 %, although lacking in the local vegetation releves. The highest Acer percentage value is found in sample 5. Single pollen grains were found in sample 2, 4, 5, 7, 9 and 11, whereas Acer turkestanica was recorded in the vegetational plots 2, 4, 7, 8, and 11. Tree species of Crataegus turkestanica and Crataegus ssp., Malus kirghisorum, Prunus sogdiana, and Cerasus mahaleb were found in the all the vegetational plots, except in plot 1, the corresponding Prunus type and Sorbus type pollen is found with scattered finds throughout the pollen samples and with single finds in samples 4, 5 and 11, respectively. Hippophaë rhamnoides reaches values of 5 % in plot 4. Except for sample 1, Artemisia shows constant values around 10 % throughout the pollen diagram. The pollen of Chenopodiaceae is represented with values of 8 – 15 % throughout the diagram. The highest percentage value of Chenopodiaceae is recorded in sample 8 where they reach 25 %. The pollen of Poaceae consistently shows values of 5 – 10 %. A slightly higher value (18 %) is recorded in sample 10. The pollen of the Asteraceae (Asteroideae and Cichorioideae type) show values up to 5 % in samples 2, 8 and 10 probably reflecting grazed meadows. Very high values of Cyperaceae are found in samples 8 and 10.

Discussion The different taxa growing in the Kyrgyz forest belts are represented very heterogeneously in the pollen samples. Some taxa such as Picea are well-represented, whereas others such as Acer are almost completely missing in the pollen, though present in the local vegetation. Similar observations were made in previous studies from other areas of the world (Wright, 1967; Wright et al., 1967) and emphasise that pollen does not mirror vegetation in a linear way (Wright, 1967; Prentice, 1985; 1986, Faegri and Iversen; 1989, Moore et al., 1991). Best fits between vegetation and pollen were obtained where the dominant tree taxa are wind- dispersed pollen producers such as Picea. In the case where insect-pollinated arboreal taxa were co-dominant or relevant (e.g. Rosaceae, Acer) the linkage is blurred, probably due to the smaller quantities of pollen produced and to the selectivity of the dispersal vector. Considering the huge differences between the taxa and vegetational belts, we discuss the relationship between plant occurrence and pollen individually for each relevant taxon.

Juniperus ssp: The pollen values suggest that Juniperus pollen follows more or less the vegetation cover of Juniperus trees at the local level (Fig 2; 4). Pollen percentage values from of 65 to 80 % may indicate mono-specific dense stands of Juniperus. At Kichiköl (KKO2, Fig. 4) values as low as 35 % indicate a crown density of 75 %. Values of 15 to 30 % may reflect local stands of individual Juniperus trees and values around 10 % may represent the pollen load of extra-local stands in a distance of 300 to 500 m. Finds of stomata corroborate the closeness of the Juniperus stands as they are only found regularly near the trees. The interpretation of Juniperus pollen percentages in mixed stands is less straightforward. Juniperus pollen percentages may be lowered by other pollen types as in the case of Karaköl (Fig. 5) and Bakaly N-S transect (Fig. 7) or they may reach high percentage values without being present in the local vegetational plots as seen at Bakaly, W-E transect (Fig. 8). Calculations of relative pollen productivity estimates (PPE) yielded a PPE higher than 3 when Poaceae were taken as reference taxon (Broström et al, 2004). Our results confirm that also the Central Asian Juniperus species display an average pollen production.

6 Picea schrenkiana: The Picea pollen seems to represent the Picea schrenkiana forest accurately. In the almost mono-specific stands around Karaköl (Fig. 5) the diminishing tree cover is correctly mirrored by reduced pollen percentages, 80 % pollen values representing a 100 % crown density and 50% pollen percentage values a 50 % forest cover. However, the plots with meadows and no Picea trees still show a Picea pollen load of 30 %. This gives us an estimate of the background value of the local Picea stands in the vicinity at the given site. Again, stomata finds corroborate the closeness and density of the Picea stands. Minor values of Picea pollen (< 10 %) in the mixed forests around Bakaly (Fig. 7; 8) reflect Picea trees in the vicinity and form a low background signal. From several studies of pollen and macrofossils it is known that Picea abies, a close relative of Picea schrenkiana can be strongly underrepresented in the pollen rain as it produces less pollen than pine, birch, alder, or hazel but that on the other hand it can be over-represented as well, due to long distance transport, different sources of contamination or inadequate coring methods (Latałowa and van der Knaap). Janssen (1966) calculated R values for several arboreal taxa at different sites in Minnesota and found that the pollen of Picea is moderately under-represented in the pollen rain.

Juglans regia: In the closed stands around Nischnie and Vierschnie lake (Fig. 9) Juglans pollen can reach percentage values that reflect the closeness of the forest. Up to 85 % were recorded in a closed mono-specific stand. The anemophilous pollen is produced in rich amounts (Zoller, 1981) and our results suggest that the pollen is transported rather well, as it is found in relatively high amounts in samples where the tree was not present locally. At Bakaly (Fig. 7; 8), where Juglans regia is at its uppermost distributional limit, less abundant pollen is found. Pollen values as low as 1 – 2 % data may reflect the scattered distribution of the Juglans trees in the local to extralocal vegetation.

Acer turkestanica: Acer pollen apparently does not reflect the occurrence of the species in the vegetation accurately. Pollen is present where no Acer tree occurred in the vegetational survey and vice versa. Generally, only single pollen grains were found. It is generally known, that insect-pollinated trees like Acer and Rosaceae are strongly under-represented in the pollen rain (Janssen, 1967; Wright et al., 1967; Faegri and Iversen, 1989). Very low Acer percentages in pollen spectra may therefore point to an appreciable share of this tree in the forest vegetation (Wright et al., 1967).

Rosaceae: The Sorbus type and Prunus type were distinguished but could not yet be assigned to the various tree species of the Rosaceae family. It is striking to see that the trees play a considerable role in the vegetation but their pollen only occurs very rarely in the surface samples. For further discussion of the Sorbus type see Faegri and Iversen (1989).

Artemisia: Though not regularly present in the local vegetational releves around the lakes, Artemisia is an important pollen type in all the diagrams. Its pollen shows high values up to 55 % in the region of Kichiköl (Fig. 2; 4), increasing with the diminishing relevance of Juniperus pollen, i.e., increasing with landscape openness. At Karaköl (Fig. 5) and at Bakaly (Fig. 7; 8) percentage values around 10 – 20 % are found. Percentage values around 10 – 20 % therefore may already reflect the background signal of the Artemisia steppes of the dry lowlands beyond the relevant source area of pollen. The high pollen values of Kichiköl probably also reflect the local regular occurrence of the taxon in the meadows near the lake, although the plant was not co-dominant there.

Chenopodiaceae: The pollen of Chenopodiaceae is present with rather low but constant values throughout the different transects. Since not regularly recorded in the vegetation releves

7 around the lakes its pollen signal may primarily reflect the background pollen loading of the Artemisia – Chenopodiaceae steppes of the lowlands. Our results support Liu et al. (1999) which found that Artemisia and Chenopodiaceae both are overrepresented in the pollen spectra in Inner Mongolia and in the Western Iran. High Artemisia pollen value can indicate the existence of steppe, but the dominant species is not certain in regions where grass- dominated steppe co-occur as Poaceae pollen is underrepresented in the pollen spectra (Wright et al., 1967; Liu et al., 1999). In the Middle East El-Moslimany (1990) has shown that Artemisia dominates in the steppe vegetation in the semi-arid zone while Chenopodiaceae and Plantago dominate the arid zone. Both Artemisia and Chenopodiaceae are characteristic of highly continental climates with cold winters and dry summers. Because Artemisia pollen increases and Chenopodiaceae decreases with decreasing aridity the ratio of their pollen (C/A) is used as a moisture indicator within a narrow geographical range and within non-forested areas.

Poaceae: At Kichiköl (Fig. 2; 4) and Karaköl (Fig. 5) the pollen of Poaceae reaches 10 – 15 % in the meadows although dominant in the herbaceous vegetation. Grass pollen is clearly underrepresented in this type of vegetation and probably masked by the high pollen representation of Artemisia and Chenopodiaceae. At Bakaly (Fig. 7; 8) Poaceae pollen reaches 30 and 55 % at its maximum therefore providing strong evidence of the low crown density (30 – 40 %) on the spot. Poaceae pollen is consistently underrepresented in the percentage values (Wright et al., 1967; Liu et al., 1999). Broström et al. (2004) argues that landscape openness may be underestimated as herb taxa (including the Poaceae) produce 6 – 8 times less pollen than tree taxa.

Cyperaceae: The pollen of Cyperaceae are mostly found in high amounts closest to the lake shore, therefore mostly reflecting the wetland vegetation around the lake. But species belonging to the Cyperaceae family are also found in meadows and to some extent in the undergrowth of forests. As this pollen type cannot be broken down to the genus or species level it is difficult to interpret the pollen values and we therefore exclude it from the pollen sum.

Sum of AP: The sum of arboreal pollen closely follows the curve of the dominant tree species in quasi mono-specific forest types: Kichiköl, Juniperus forest (Fig. 2; 4), Karaköl, Picea forest (Fig. 5) and Nischnie, Vierschnie Ozira, Juglans forest (Fig. 9). The tree crown cover explains 76 and 85 % of the AP variability in the Juniperus forest and in the Picea forests, respectively (Fig. 3; 6). In the Juglans forest and in the mixed forest at Bakaly only 35 and 32 % (Fig 10, Fig. 11) are achieved, although the correlation coefficients are still rather high (r = 0.59 and 0.57). Vegetation patterning can affect the pollen loading (Bunting et al., 2004) and several factors, such as species composition, spatial pattern and structure of the vegetation influence the spatial scale of provenance of pollen (Sugita et al., 1999). As trees with good pollen production (Juglans, Picea and Juniperus) are interspersed with low pollen producers (Acer, Rosaceae, Pistacia) these factors may come to play a more important role. If all forest types are pooled together, 44 % of the variability of AP is explainable by the crown density (r = 0.66, Fig. 12). We therefore conclude that AP percentages are good proxy to estimate the tree crown cover on the spot.

Conclusions Quantitative reconstruction of past and present plant abundance and distribution from fossil and subfossil pollen records have been the goal of palynological research since the earliest days yet remain elusive (Bunting and Middleton, 2005). Empirical (Janssen, 1967; Calcote

8 1995; Jackson and Kearsley, 1998; Liu et al., 1999; Shen et al., 2006 and many others) as well as modelling studies (Sugita 1994, Sugita et al., 1999, and all the others of the group around POLLANDCAL) have shown that the relationship between vegetation and pollen imprint is not straightforward but is affected by pollen productivity, pollen dispersal, vegetation patterning and species composition (Sugita et al., 1999, Bunting et al., 2004). The present work will add up to the body of knowledge that has been carried together and will help to better reconstruct the Quaternary history of Kyrgyzstan and Central Asia by means of fossil pollen records.

Acknowlegdments We thank Brigitta Ammann, Marc Lehmann, Zakir Sarymsakov, Willi Tanner, Bronislav I. Venglovsky, and all Kyrgyz friends who allowed the extensive field trips in 2003 and 2005. We are very grateful to Intercooperation, Project Kirfor (Kyrgyz-Swiss Forestry Support Programme), to the University of Bern, to the Hochschulstiftung, and to scnat (Akademie der Naturwissenschaften Schweiz) for financing and supporting this study.

Literature Beug, H.-J. 2004: Leitfaden der Pollenbestimmung für Mitteleuropa und angrenzende Gebiete. Dr. Friedrich Pfeil Verlag, München.

Blaser, J., Carter, J. and Gilmour, D., editors, 1998: Biodiversity and sustainable use of Kyrgyzstan’s walnut-fruit forests. English version. IUCN, Intercooperation, Gland, Cambridge, Bern.

Broström, A., Sugita, S. and Gaillard, M.-J. 2004: Pollen productivity estimates for the reconstruction of past vegetation cover in the cultural landscape of southern Sweden. Holocene 14, 368 – 81

Broström, A., Sugita, S., Gaillard, M.-J. and Pilesjö, P. 2005: Estimating the spatial scale of pollen dispersal in the cultural landscape of southern Sweden. Holocene 15, 252 – 62

Bunting, M.J., Gaillard M.-J., Sugita, S., Middleton R. and Broström, A. 2004: Vegetation structure and pollen source area. Holocene 14, 651 – 60

Bunting M.J. and Middleton D. 2005: Modelling pollen dispersal and deposition using HUMPOL software, including simulating windroses and irregular lakes. Review of Palaeobotany and Palynology 134, 185 – 196

Calcote, R. 1995: Pollen source area and pollen productivity: evidence from forest hollows. Journal of Ecology 83, 591 – 602

El-Moslimany, A.P. 1990: Ecological significance of common nonarboreal pollen: examples from drylands of the Middle East. Review of Palaeobotany and Palynology 4, 343 – 50

Epple, C. 2001: A vegetation study in the walnut and fruit-tree forests of Southern Kyrgyzstan. Phytocoenologia 31, 571 – 604

Faegri, K. and Iversen, J. 1989: Textbook of Pollen Analysis 4th edition. John Wiley & Sons.

9

Firbas, F. 1934: Über die Bestimmung der Walddichte und der Vegetation waldloser Gebiete mit Hilfe der Pollenanalyse. Planta 22, 109 – 145

Jackson, S.T. and Kearsley, J.B. 1998: Quantitative representation of local forest composition in forest-floor pollen assemblages. Journal of Ecology 86, 474 – 90

Janssen, C. R. 1967: A comparison between the recent regional pollen rain and the sub- recent vegetation in four major vegeation types in Minnesota (U.S.A.). Review of Palaeobotany and Palynology 2, 331 – 42

Kolov, O. 1998: Ecological characteristics of the walnut-fruit forests of southern Kyrgyzstan. In Blaser, J., Carter, J. and Gilmour, D., editors: Biodiversity and sustainable use of Kyrgyzstan’s walnut-fruit forests. English version. IUCN, Intercooperation, Gland, Cambridge, Bern, 59 – 61

Kvavadze, E.V. and Stuchlik, L. 1990: Subrecent spore-pollen spectra and their relation to recent vegetation belts in Abkhazia (north-western Georgia, USSR). Acta Palaeobotanica 30, 227 - 57

Latałowa, M. and van der Knaap, W.O. 2006: Late Quaternary expansion of Norway spruce Picea abies (L.) Karst. in Europe according to pollen data. In press.

Liu, H., Cui, H., Pott, R. and Speier, M. 1999: The surface pollen of the woodland-steppe ecotone in southeastern Inner Mongolia, China. Review of Palaeobotany and Palynology 105, 237 – 50

Moore, P.D., Webb, J.A. and Collinson, M.E. 1991: Pollen analysis. Blackwell scientific publications, Oxford.

Prentice, I. C. 1985: Pollen representation, source area and basin size: toward a unified theory of pollen analysis. Quaternary Research 23, 76 – 86

Prentice, I. C. 1986: Forest-composition calibration of pollen data. In Berglund, B.E., editor, Handbook of Holocene Palaeoecology and Palaeohydrology. John Wiley & Sons Ltd, 799 – 817

Reille, M. 1992: Pollen et spores d’Europe et d’Afrique du Nord. Laboratoire de Botanique Historique et Palynologie, Marseille.

Reille, M. 1998: Pollen et spores d’Europe et d’Afrique du Nord. Supplément 2. Laboratoire de Botanique Historique et Palynologie, Marseille.

Shen, C., Kam-biu, L., Lingyu T. and Overpeck, J.T. 2006: Quantitative relationships between modern pollen rain and climate in the Tibetan Plateau. Review of Palaeobotany and Palynology 140, 61 – 77

Sugita, S. 1994: Pollen representation of vegetation in Quaternary sediments: theory and method in patchy vegetation. Journal of Ecology 82, 881 – 97

10 Sugita, S., Gaillard, M.-J. and Broström, A. 1999: Landscape openness and pollen records: a simulation approach. Holocene 9, 409 – 21

Venglovsky, B.I. 1998: Potentials and constraints for the development of the walnut-fruit forests of Kyrgyzstan. In Blaser, J., Carter, J. and Gilmour, D., editors, Biodiversity and sustainable use of Kyrgyzstan’s walnut-fruit forests. English version. IUCN, Intercooperation, Gland, Cambridge, Bern. 73 – 6

Welten, M. 1950: Beobachtungen über den rezenten Pollenniederschlag in alpiner Vegetation. Sonderdruck aus E. Rübel und W. Lüdi, Bericht über das Geobotanische Forschungsinstitut Rübel in Zürich für das Jahr 1949.

Wright Jr., H.E. 1967: The use of surface samples in Quaternary pollen analysis. Review of Palaeobotany and Palynology 2, 321 – 30

Wright Jr., H.E., McAndrews J.H. and van Zeist, W. 1967: Modern pollen rain in Western Iran and its relation to plant geography and Quaternary vegetational history. Journal of Ecology 55, 415 - 43

Zoller, H. 1981: Pteridophyta, Spermatophyta. In Hegi, G: Illustrierte Flora von Mitteleuropa, Band III Angiospermae, Dicotyledones 1, Teil 1. Verlag Paul Parey, Berlin, Hamburg. 3 – 15

11 Figure captions: Table 1: Vegetation types and tree crown density of pollen surveys

Fig 1: map of Kyrgyzstan and the four studied sites (Kichiköl, Karaköl, Bakaly, Nischnie and Vierschnie Ozira) Fig. 2: pollen percentages of selected pollen types at Kichiköl, sampled along a gradient of crown densities in a Juniperus forest. Crown density in sample 1 is 10 %, in sample 2: 20 %, in sample 3: 45 %, in sample 4: 80 % and in sample 5: 100 % Fig. 3: Correlation between tree cover and pollen percentages of arboreal pollen (AP) in the Juniperus forest at Kichiköl. Fig. 4: Vegetation transect and pollen percentages of selected pollen types at Kichiköl, sampled along a W-E transect. Crown density of Juniperus in sample 1 is 75 %, in sample 2: 30 %, in sample 3 to 5 no tree cover is recorded. Legend: symbols of the tree types used in the figures Fig.5: Vegetation transect and pollen percentages of selected pollen types at Karaköl, sampled along a N-S transect. Crown density of Picea schrenkiana in sample 1 is 100 %, in sample 2: 50 %, in sample 3: 0 %, in sample 4: 25 %, in sample 5: 75 % and in sample 6: 100 %. Fig. 6: Correlation between tree cover and pollen percentages of AP in the Picea forest at Karaköl. Fig 7: Vegetation transect and pollen percentages of selected pollen types at Bakaly, N-S transect. Tree crown cover in sample 1 is 70 %, in sample 2 and 3: 0 %, in sample 4: 50 %, in sample 5: 70 % and in sample 6: 80 % Fig. 8: Vegetation transect and pollen percentages of selected pollen types at Bakaly, W-E transect. Tree crown cover in sample 1 is 60 %, in sample 2: 0 %, in sample 3: 50 %, in sample 4: 80 %, in sample 5: 0 %, in sample 6: 40 %, in sample 7: 0 %, in sample 8: 40 %, in sample 9: 30 % and in sample 10: 40 %. Fig. 9: Vegetation transect and pollen percentages of selected pollen types at Nischnie and Vierschnie Ozira. Tree crown cover in sample 1 is 50 %, in sample 2: 30 %, in sample 3: 60 %, in sample 4: 30 %, in sample 5: 40 %, in sample 6: 80 %, in sample 7: 40 %, in sample 8: 50 %, in sample 9: 60 %, in sample 10: 40 % and in sample 11: 60 % Fig. 10: Correlation between tree cover and pollen percentages of arboreal pollen in the Juglans forest at Nischnie and Vierschnie Ozira. Fig. 11: Correlation between tree cover and pollen percentages of arboreal pollen in the mixed forest at Bakaly. Fig. 12: Correlation between tree cover and pollen percentages of arboreal pollen of all four sites in Kyrgyzstan.

12 Table 1: Vegetation types of pollen surveys

Juniperus forest, Juglans forest Picea forest, Mixed forest sample TCC sample TCC locality number dominant plant species (%) locality number dominant plant species (%) KKO1 1 Juniperus turkestanica, J. semiglobosa 10 KAO 1 Picea schrenkiana 100 2 Juniperus turkestanica, J. semiglobosa 20 2 Picea schrenkiana 50 3 Juniperus turkestanica, J. semiglobosa 45 3 Poaceae, upland herbs 0 4 Juniperus turkestanica, J. semiglobosa 80 4 Picea schrenkiana, Juniperus sp. 25 5 Juniperus turkestanica, J. semiglobosa 100 5 Picea schrenkiana 75 KKO2 1 Juniperus turkestanica, J. semiglobosa 75 6 Picea schrenkiana 100 2 Juniperus turkestanica, J. semiglobosa 30 BAO2 1 Crataegus turkestanica, Malus kirghisorum, Juniperus turkestanica, J. 70 semiglobosa, Cerasus mahaleb 3 Artemisia sp., Poaceae, upland herbs 0 2 Exochorda tianschanica, Lonicera nummularifolia, Malus kirghisorum 0 4 Artemisia sp., Poaceae, upland herbs 0 3 Cyperaceae 0 5 Artemisia sp., Poaceae, upland herbs 0 4 Malus kirghisorum, Crataegus turkestanica 50 NOVO 1 Juglans regia 50 5 Picea schrenkiana, Crataegus turkestanica 70 2 Acer turkestanica, Crataegus turkestanica, Malus kirghisorum, Prunus 30 6 Picea schrenkiana, Acer turkestanica, Cerasus mahaleb, Juniperus 80 sogdiana turkestanica 3 Juglans regia 60 BAO 1 Juglans regia, Cerasus mahaleb, Acer turkestanica, Juniperus turkestanica 60 4 Acer turkestanica, Crataegus turkestanica, Malus kirghisorum, Prunus 30 2 Cotoneaster multiflora, Rubus caesus, Spirea hypericifolia 0 mahaleb 5 Cerasus mahaleb, Juglans regia, Prunus sogdiana 40 3 Juniperus turkestanica, Acer turkestanica, Crataegus turkestanica 50 6 Juglans regia, Cerasus mahaleb, Crataegus sp., Prunus sogdiana 80 4 Betula sp. 80 7 Acer turkestanica, Juglans regia, Crataegus sp. 40 5 Cyperaceae 0 8 Acer turkestanica, Crataegus turkestanica, Malus kirghisorum 50 6 Juniperus turkestanica, J. semiglobosa, Malus kirghisorum 40 9 Cerasus mahaleb, Crataegus turkestanica, Juglans regia, Malus 60 7 Exochorda tianschanica, Berberis heteropoda 0 kirghisorum, Prunus sogdiana 10 Crataegus sp., Malus kirghisorum 40 8 Acer turkestanica, Juniperus turkestanica, Juglans regia, Malus kirghisorum 40 11 Acer turkestanica, Crataegus turkestanica, Juglans regia, Malus 60 9 Juniperus turkestanica, Juglans regia 30 kirghisorum, Prunus sogdiana 10 Juniperus turkestanica, J. semiglobosa, Malus kirghisorum 40

TCC: Tree crown cover

13

Fig 1

14

Fig 2

15

Fig. 3

16

Fig 4

17

Legend attached to figure 4

18

Fig 5

19

Fig. 6

20

Fig. 7

21

Fig. 8

22

Fig. 9

23

Fig. 10

Fig. 11

24

Fig. 12

25